Clinical charge profiles prediction for patients diagnosed with chronic diseases using Multi-level Support Vector Machine

Frequent Emergency Department Utilization and Behavioral Health Diagnoses

“…Multiple algorithms have been developed to measure disease complexity and comorbidity for practice and research (Guralnik, 1996), but most focus on hospitalized cases to evaluate mortality (Charlson, Pompei, Ales, & MacKenzie, 1987;Grendar et al, 2012;Naessens, Leibson, Krishan, & Ballard, 1992) and hospital length of stay (Young, Kohler, & Kowalski, 1994;Zhong, Chow, & He, 2012), rather than incorporate diagnoses from community care settings to represent the population at large. Multiple algorithms have been developed to measure disease complexity and comorbidity for practice and research (Guralnik, 1996), but most focus on hospitalized cases to evaluate mortality (Charlson, Pompei, Ales, & MacKenzie, 1987;Grendar et al, 2012;Naessens, Leibson, Krishan, & Ballard, 1992) and hospital length of stay (Young, Kohler, & Kowalski, 1994;Zhong, Chow, & He, 2012), rather than incorporate diagnoses from community care settings to represent the population at large.…”

Data mining and predictive analytics applications for the delivery of healthcare services: a systematic literature review

“…There have been 7 studies grouped under the main HOSCM function area whereas these 8 studies can be further subdivided in to strategic and medium to short term healthcare resource planning, outpatient scheduling and Queuing Analyses for workforce scheduling. The strategic healthcare resource planning has been on a macro scale with studies estimating cost of stay funding policies for at national, state and local levels (Zhong et al, 2012) and planning for public health availability and accessibility (Lavrač et al, 2007). Ng et al (2006) used data mining for early prediction of patients requiring longer hospital care, therefore, contributing to short to medium term hospital resources planning for inpatient care.…”

“…This category was used to collect and collate information regarding the data sources, sample size, data description and observation period (Table 3). Most studies used the information stored in the electronic health records (EHRs) of hospitals/medical centres whereas in a few cases such as Zhong et al (2012) used dataset on United States healthcare cost & utilization to predict clinical charge profiles, Chi et al (2008) used the Iowa state Inpatient dataset (SID) for their expert referral system and Rubrichi and Quaglini (2012) used Italian national pharmacy database for their text mining project. There also have been two instances where along with the EHRs data, actual data was also collected for the project.…”

Analysis of diabetic patients through their examination history

“…Concerning the analysis of medical data for diabetic patients, several works, mainly exploiting classification techniques, have been proposed (Karegowda et al, 2012;Meng et al, 2012;Mohamudally & Khan, 2011;Zhong et al, 2012). Classification is a supervised data mining approach that assigns new unlabeled data to a class label by means of a model built from data with known class label (Pang-Ning T. and Steinbach M. and Kumar V., 2006).…”

“…This work showed that the decision tree model (C5.0 (Pang-Ning T. and Steinbach M. and Kumar V., 2006)) yielded the best accuracy, followed by logistic regression, and artificial neural networks. The work in (Zhong et al, 2012) proposed a multi-level support vector machine approach to classify and predict clinical charge profiles as well as the length of hospital stay for patients affected by heart, diabetes, and cancer diseases. In (Mohamudally & Khan, 2011), different data mining algorithms (the K-means algorithm, C4.5 decision tree classifier, artificial neural networks, and 2D graphs for data visualization) are integrated to predict, classify, and visualize a medical diabetic dataset.…”